The RRD file stores data in 4 different granularities:
-- 10 second average per point x 360 = 1 hour of data
-- 60 second average per point x 360 = 6 hours of data
-- 120 second average per point x 360 = 12 hours of data
-- 180 second averages per point x 480 = 24 hours of data per point
So you can select how fine the data is, but it limits the total collection time of the data. The CSV file gives you the same interval that the graph would, which is based on how much data is in the file. You can get the different ranges from the CSV exporter by adding "&nancnt=X" to the end of the url.
-- &nancnt=460 is 1 hour
-- &nancnt=360 is 6 hours
-- &nancnt=240 is 12 hours
-- &nancnt=0 is 24 hours
These numbers come from the drop down box for range under the graph on the homepage and override the auto-ranger's count of blank data (nan) in the data file. If you're looking for 10 second averages over 24 hours or something like that, we don't store that much data, just enough data to fill the graph.